An Algorithm Design for Fish Recognition Based on Computer Vision

Fish image recognition has always been a challenging task. In this paper, we propose a recognition algorithm based on computer vision for the four major species, silver carp, bighead carp, herring and grass carp. Firstly, the algorithm uses guided filtering to filter out noise and blur points in the sample image, which can help obtain accurate contour feature values of the fish body. Then the image is enhanced in order to avoid the loss of the contour of the fish body due to the influence of brightness. After the contour of the fish is drawn, the feature points on the contour are extracted according to fishtail angle and aspect ratio, based on which the fish kind could be concluded. The research results show that the recognition algorithm provided in this study can accurately classify and identify four kinds of fish with an accuracy of 96.25%.